On Sun, 2006-05-07 at 14:32, Mikkel L. Ellertson wrote:
>
Well, if you are using dd to copy a working disk image, you are
going to have to make more changes anyway. You may have problems
with the host name, depending on how it is set.
Normally I clone a machine set for DHCP so it can be
brought up anywhere and an appropriate name address
assigned later (the destination may not be known at
the time it is copied).
You may also have a
problem in that the MAC address of the NIC will not match the one in
the ifcfg-eth? script and/or /etc/iftab (Depending on the
distribution.)
Yes, burying the MAC address in an obscure config file is
a bad idea as well, and the way it was handled could have
caused me some serious trouble if I hadn't caught on
by accident. I've forgotten the version and update level
but at some point the network-config program started
squirreling away the MAC address but not paying much attention
to it. So, I deployed a bunch of machines where the
last step after cloning was to boot the disk locally
to set the hostname and IP address, then the SCA disks
and carriers were shipped to remote locations and swapped
into the machines where they were intended to run. Everything
worked fine, time passed, updates happened, and then after some
update, a machine falls off the network. Turns out that the
update changed the init script behavior so that those config
files with mismatching MAC addresses were ignored.
Fun stuff... And of course I couldn't duplicate the problem
on my local box which was an exact copy of what was deployed.
I was just lucky that the first one updated was at a nearby
location and I went over and hooked up a monitor to find that
refusing to start the network was intentional.
The use of labels was started because there were too
many newbies that would move the drive, and then not be able to boot
it. One common problem was where they would have two drives, in the
machine, one with Windows and one with Linux with Linux on a drive
other then the Windows drive. Then they would decide to move the
Linux drive to its own machine, and it would not work. When using
labels, all you have to do is re-install the boot loader, and maybe
tweak the config for it. When using the actual devices, you have to
edit /etc/fstab as well.
So as long as you only ever have one fedora/rh/centos installation
this won't break anything. What was the plan for later, when
people have more than one drive and connect them together?
I feel that the default install should make things as easy as
possible for a newbie, doing a normal install, and have the advanced
options for when you are not doing a standard install.
What is it about refusing to boot when duplicate labels are
seen that you think is newbie-friendly? An option or default
to continue, using one or the other would be friendly. Crashing
is not friendly.
If you are
not doing a standard install, expect to have to use some of the
advanced options. Installing on one machine, and cloning the install
is NOT a standard install!
It could be. There is nothing about making the machine
cloneable that makes it less useful or handy in the initial
install. Everyone, even a newbie, should be able to back up
and restore their machine which is essentially the same thing,
especially if you restore to different hardware.
Now, one change that would help is the way duplicate labels are
handled. Right now, the last label found of the duplicates found is
used.
Not unless it has recently changed. I've never been able to
boot a machine with duplicate root labels.
--
Les Mikesell
lesmikesell(a)gmail.com